USB DAC Design questions

[ble0t]

Quote:

Originally Posted by Clutz People will need this flexibility because people will probably want to be able to play different audio formats through their computer (i.e. other than CD Audio)?

This is where the ASRC comes in...as Francis said, it should solve the issue of multiple input frequencies. Also, if I'm not mistaken, the ASRC helps to reduce jitter better than a PLL (see this thread over at diyaudio).

 

[ble0t]

Also, I like the idea that 00940 had in his last post in the older thread about using something like the PCM2707 for USB and then also having a CS8416 available for SPDIF/Optical/AES (using transformers where needed). Having each section (DIR/DAC/Analog) on its own board it also a great idea. How about shooting for something like this?

 

[00940]

Quote:

Originally Posted by guzzler There is the other option of using the 1793 which has differential voltage outputs. Aos used that in his Ally and Piccollo DACs. It's not got such a good reputation as the 1794/8, but just thought I'd mention it.

In the same class as the 1793, with voltage out, we could also consider the wolfson wm8740, used in the cambridge 640c, or the CS4398, used in the headroom microdac. Still, the ability to tailor your own I/V stage leads my heart towards dac such as the pcm1794/98.

Francis : some designs using the pcm2707 and direct I2S to pcm1798have been posted earlier. http://www6.head-fi.org/forums/showthread.php?t=134125 for example. I agree that an asrc is better, but is it really a necessity ?

A local clock can get from cheap to very expensive. Considering the budget, we cannot afford something as a tent clock. We can of course the ad1896 with only a crystal iirc.

If we add a cs8416, transformers, and so on, costs will jump up.

Btw, the old schematics I posted are far from perfect. They're missing inductors on the power lines for example.

 

[Clutz]

Quote:

Originally Posted by 00940 In the same class as the 1793, with voltage out, we could also consider the wolfson wm8740, used in the cambridge 640c, or the CS4398, used in the headroom microdac. Still, the ability to tailor your own I/V stage leads my heart towards dac such as the pcm1794/98. Francis : some designs using the pcm2707 and direct I2S to pcm1798have been posted earlier. http://www6.head-fi.org/forums/showthread.php?t=134125 for example. I agree that an asrc is better, but is it really a necessity ? A local clock can get from cheap to very expensive. Considering the budget, we cannot afford something as a tent clock. We can of course the ad1896 with only a crystal iirc. If we add a cs8416, transformers, and so on, costs will jump up. Btw, the old schematics I posted are far from perfect. They're missing inductors on the power lines for example.

My personal opinion on the matter is that if we end up trying to include SPDIF inputs, and thus need a cs8416, then we either have to up the budget substantially, or make compromises elsewhere. My goal out of this was to make a pretty good quality USB only DAC. I'm not saying I'm against it these other things, I just think we need to weigh the costs and benefits. Towards this end, my preference, I think, would be to put money towards getting a better clock, rather than multiple inputs.. but of course, I'm only one person.

Re the ASRC: Is the principle benefit that if we use it, then we are not constrained to 16bit @ 44.1kHz i2s data streams into the DAC? Because if it is, I think we have a problem further up the food chain. The PCm2707 only provides a 16bit @ 44.1khz i2s data stream (or am I wrong?)? So would we be connecting the PCM2707's SPDIF output to the ASRC?

 

[ble0t]

00940: I see your point. I think the idea of having seperate boards for the various aspects of the DAC would work well in this situation. I guess my thought would be to come up with a single DAC section since this should be the most standard part and then possibly differentiate the DIR section and Analog section from there. Although not completely necessary, I like the idea of having an ASRC in there if possible. Total extra cost (if using a standard crystal) would only be about $20 extra, which IMO would be worth it.

 

[steinchen]

I'm very interested in an DIY DAC, too. Since I already got a Piccolo DAC I'm interested in a high end DAC only.

I advocate a modular design with input section, DAC and I/V section. The DAC section could be replaced by a board with a new DAC chip after some years. Concerning the I/V section I expect many diverging opinions and requirements (single ended vs balanced, passive vs opamp vs discrete).

 

[guzzler]

I would argue against a modular system. As F_V mentions too, we're dealing with fast MHz signals, and to acheive a good interfacing between the separate parts requires a good deal of thought. Also, it considerably worsens the board layout going for "flexibility" (ie, people want a board without designing it themselves). You could simply put a couple of pads on the Iout pins for people to run to their board should they so wish

Set a definite limit on the design parameters and stick with them. There's always tradeoffs, so better to live with them than to worry about them.

 

[Clutz]

So then, how should we decide what the design parameters are? One important design parameter to take into consider is the budget, but that can mean many things. Is it better to go with a simpler design and build the budget assuming more expensive parts (boutique caps, fancier resistors), or assume that use good quality parts (i.e. panasonic FMs, bc components resistors, maybe vishays)?

My design parameters would be a USB only DAC. I personally would be happy with something like the PCM2707->ASRC->PCM1794->OPA1632, or PCM2707->PCM1794->OPA1632, and possibly using some of the savings to offset a better clock? It would be mains powered- at least two seperate transformers, one with dual 5 volt secondaries to make our 5v and 3v3, and one with dual 12 volt secondaries to power the op-amps with, and possibly take the seond 12 volt tap and drop it also down to 5 if we wanted another 5 volt PSU?

I'd imagine we want something better than a TREAD for the PSU- maybe something like a stripped STEPS a la dsavitsk's nonos DAC's PSU?

 

[ble0t]

After thinking it through, it makes sense to go all one board with this as guzzler said. Modularity would be nice, but it probably makes itself impractical. I guess my vote would be for PCM2707 --> (ASRC or some type of reclock) --> PCM1798 (big cost savings over PCM1794 and still very good performance as well as diff out) --> I/V Stage. As far as the power supply is concerned, it would be possible to include it onboard if making use of chips like the reg101 and lt1764/lt1964 for the various stages.

 

[dsavitsk]

Quote:

Originally Posted by Clutz maybe something like a stripped STEPS a la dsavitsk's nonos DAC's PSU?

That's the Baby Steps (TM) to you.

The problem with this design, as I have found out, is that since some of the filtering is pre transformer and some is post transformer, this requires the transformer being close to the rest of the circuit. The irony is that this adds more noise than the circuit cleans, so I think it is a net loss. This argues for making the PS on a seperate PCB or leaving out the pre transformer filters and assuming that anyone who wants them will add them seperatly.

Quote:

Originally Posted by steinchen Concerning the I/V section I expect many diverging opinions and requirements (single ended vs balanced, passive vs opamp vs discrete).

Don't forget hollow state options, which would need a modular board. I think this is a good place for modularity -- perhaps picking one option to put on the board, but also adding Iout pads for others.

 

[00940]

-The power supply should be onboard, but not the xformers imo. CRCRC filters followed by reg101 and the like should be perfect and compact.

- The difference in between pcm1794 and pcm1798 is mainly in the digital filter. If we arrive with a 96khz or higher signal out of an ad1896, the resampling inside the dac chip will happen further away from the signal band (see the posts in my old thread). The pcm1798 could thus be sufficient.

- For a DAC, the use of SMD parts makes a lot of sense (for about anything but the electrolytics). It will avoid the overuse of expensive boutique stuff.

 

[Clutz]

Yes, the PCM1798 is quite a bit cheaper than the 94 at digikey, about $10 less. That said, I think that the PCM1794 and 1798 are pin compatible, so if someone wanted to use the 1794 instead of the 98, they would be able to do so, so we can go with either of them. But I would agree that on the budget front, we should do the costing of the project based on the 1798.

 

[Clutz]

Quote:

Originally Posted by 00940 -The power supply should be onboard, but not the xformers imo. CRCRC filters followed by reg101 and the like should be perfect and compact. - For a DAC, the use of SMD parts makes a lot of sense (for about anything but the electrolytics). It will avoid the overuse of expensive boutique stuff.

Makes a lot of sense to me.

Quote:

Originally Posted by dsavitsk Don't forget hollow state options, which would need a modular board. I think this is a good place for modularity -- perhaps picking one option to put on the board, but also adding Iout pads for others.

I really think that the Iout pads are a good idea.

 

[ble0t]

Sounds good to me. I would be willing to assist on the schematic/layout. What layout/PCB prog did you use for your original design 00940? Perhaps we could make that a starting point

 

[00940]

like everyone, eagle free version. But I'm not sure i've much left.